China has formally extended its Xinchuang technology localization campaign to artificial intelligence chips, creating a new certification category for “AI training and inference chips” and publishing the first batch of approved domestic products. As reported by the South China Morning Post, the China Information Technology Security Evaluation Center and the National Secrecy Science and Technology Evaluation Center jointly released the results on Tuesday evening, marking the first time AI processors have been formally incorporated into the country’s “secure and reliable” (anquan kekao, or Anke) technology framework. Certifications under the new category are valid for three years.
The Xinchuang initiative, whose name derives from the Chinese term for “information technology application innovation,” has been the primary mechanism through which Beijing has systematically replaced foreign hardware and software in sensitive government and state-enterprise systems. The campaign accelerated around 2020 and expanded significantly after 2022, encompassing the finance, energy, and telecommunications sectors. The first Xinchuang assessment results, covering central processing units, operating systems, and databases, were published in December 2023. The extension of the AI chip program in May 2026 represents the most strategically significant expansion of the program to date, given the centrality of AI accelerators to modern computing infrastructure.
(Related: China Bans Foreign AI Chips from State-Funded Data Centers)
The Certified Chip Roster and Notable Absences
Nine domestic chip developers received certifications in the inaugural AI chip batch. Huawei Technologies’ Ascend 310 and Ascend 910 processors were included, as were Alibaba Group Holding’s T-Head Zhenwu M530 and M890 chips. The list also encompasses graphics processing units from Biren Technology, Hygon Information Technology, Iluvatar CoreX, MetaX, and Moore Threads. All products passed the rigorous Anke V3.0 technical requirements, which assess security, reliability, and compatibility with domestic software ecosystems.
Two notable companies were absent from the inaugural list: Cambricon Technologies and Baidu-backed Kunlunxin. Participation in the Xinchuang assessment process is voluntary, and companies may choose not to submit products if their primary market focus lies outside government and state-enterprise procurement. Nevertheless, exclusion from the official catalogue limits access to a substantial and growing pool of mandated procurement contracts. The absence of Cambricon, a publicly listed AI chip designer that has historically positioned itself as a domestic Nvidia alternative, may reflect ongoing technical challenges in meeting the Anke V3.0 requirements or a strategic decision to focus on commercial cloud customers.
(Related: China’s Domestic Chipmakers Seize 41% of the Local AI Market)
Policy Convergence and Market Implications
The AI chip certification release coincides with a broader policy push to accelerate domestic adoption of home-grown semiconductors. Just last week, the National Development and Reform Commission announced plans to guide domestic large language model developers to increase their adaptation to Chinese-made chips, a directive that directly complements the Xinchuang certification framework by creating demand-side pressure alongside the supply-side certification signal.
According to an April 2026 report by JPMorgan, China’s indigenous AI chip demand measured in units could grow at a compound annual rate of approximately 40% through 2030. The investment bank identified wafer fabrication capacity as the primary bottleneck constraining faster adoption, but projected that significant shipments of locally produced chips would begin in the second half of 2026. The Xinchuang certification of nine domestic AI chip products provides these manufacturers with a formal government endorsement that should accelerate procurement decisions across the state sector, potentially pulling forward some of that anticipated H2 2026 demand. For investors tracking China’s semiconductor localization trajectory, the inaugural AI chip certification list offers the clearest signal yet of which domestic suppliers are positioned to capture the largest share of the government-mandated replacement market.
The three-year certification validity period is also worth noting. Products that pass the Anke V3.0 assessment in May 2026 will retain their certified status until approximately mid-2029, giving procurement officers a stable reference list for multi-year infrastructure planning. This duration aligns with typical government IT procurement cycles, which often span three to five years from initial specification to full deployment. For Huawei and Alibaba T-Head in particular, the certification provides a durable competitive advantage in government and state-enterprise AI infrastructure tenders, a market segment that is expected to grow substantially as Beijing accelerates its push to deploy AI across public administration, healthcare, and national security applications.
The inaugural AI chip certification list also sets a precedent for how the Xinchuang program will evolve. The voluntary nature of the assessment means that the list will expand over time as additional vendors submit products for evaluation, and future assessment cycles may introduce updated technical requirements that reflect advances in AI chip architecture. Companies that achieve early certification gain a first-mover advantage in government procurement, but they must maintain compliance with evolving Anke requirements to retain their certification. This creates an ongoing incentive for domestic AI chip developers to align their product roadmaps with the technical standards set by the certification bodies, a dynamic that effectively gives the Chinese government a degree of indirect influence over the direction of domestic AI chip development.